Assessment of Commercial Laboratories Performing Hair Mineral Analysis

As public health officials in a state health department, we were asked by a local health officer for assistance with interpreting results from hair testing performed on residents living near a hazardous waste site to assess their exposures to metals. We were surprised to learn that an average of 225 000 hair mineral tests costing $9.6 million are performed yearly by 9 laboratories in the United States (Table 1). Some of these laboratories included claims in their client brochures such as "Hair element analysis . . . can assist the physician with an early diagnosis of physiological disorders associated with aberrations in essential and toxic element metabolism."

In 1985, Barrett¹ assessed the reliability of commercial hair mineral test results and found that "the reported levels of most minerals varied considerably between identical samples." He suggested that actions taken by the Federal Trade Commission to bar 1 laboratory from making false claims would make it "unlikely that any hair analysis laboratory will advertise directly to the public again," and he encouraged additional government action to protect the public. Because laboratory methods generally have improved since 1985 and because the Clinical Laboratory Improvement Act (CLIA) was passed in 1988 to regulate clinical testing, we decided to update Barrett's results. Hypothesizing that there should be no significant variations between test results on scalp hair from a single donor, we performed a small evaluation of interlaboratory agreement among laboratories currently offering these tests.

Laboratories were identified through California local health officers, other investigators, the Internet, and the laboratories themselves. A total of 9 US commercial laboratories were identified that advertised hair mineral analysis for health assessment purposes. The number of laboratories selected was limited by the amount of scalp hair samples available from the donor. Six laboratories were selected, based first on laboratory volume of hair analyses and second on lower sample weight requirements. The 6 laboratories represent 90% of analyses performed and a cross-section of analytical methods used by all identified laboratories. Laboratory characteristics were identified from laboratory client brochures, Internet sites, and telephone interviews of laboratory personnel. In the interviews, one author (S.S.) asked the technical staff from each laboratory a list of open-ended questions about sample preparation and analysis, data interpretation, and quality assurance/quality control.

Hair was donated by 1 author, a generally healthy 40-year-old white woman, with untreated brown hair. The hair was cut adjacent to the scalp from several areas of the parietal-occipital regions, using clean stainless steel scissors. Hair cuttings measuring 3.2 cm from the scalp were combined, mixed, weighed, and divided into the amounts required by each laboratory, from 0.25 to 1.0 g. One split of the hair sample was sent to each laboratory, using the sample envelope provided by the laboratory. The procedure was designed to address laboratory criticisms of the Barrett study,¹ that is, that the hair came from more than 1 donor, it was not taken from the scalp, and it was up to 15 cm long, leading to nonhomogeneity in Barrett's pooled sample.

Each laboratory sent a report consisting of quantitative results and their interpretation of health implications. The distributions of reported values for each element were tested for extreme (outlier) values.² Values reported as "<x" were treated as x/2. Differences among laboratories, adjusting for the differences across elements, were analyzed by Friedman analysis of variance for randomized blocks,² a nonparametric test that compares the ranking of each laboratory's values across elements. Independence in a cross-classified 3 × 6 table was tested by Fisher exact test,² using a search tree algorithm implemented in Stata.³ The laboratories' interpretations of health implications were compared qualitatively.

Characteristics of all 9 identified laboratories (A-I) are summarized in Table 1. All laboratories advertised CLIA certification in their client brochures. Follow-up with the Health Care Financing Administration (HCFA), the agency that enforces CLIA, confirmed that laboratory B had misrepresented itself as CLIA-certified. For the remaining laboratories, CLIA certifications were typically in specialty areas, such as chemistry or toxicology. (There is no CLIA specialty area for hair analysis.) All laboratories required hair samples to be submitted under the signature of a health care professional legally authorized to request laboratory tests (eg, a physician or chiropractor).

Cost for hair testing ranged from $30 to $69 per sample for a typical analysis, including the laboratory's interpretation. The number of hair elements routinely analyzed ranged from 16 to 39. Nineteen elements analyzed by all laboratories, except laboratory I (Co, Ni, and P were not analyzed), included Al, As, Ca, Cd, Co, Cr, Cu, Fe, Hg, K, Mg, Mn, Mo, Na, Ni, P, Pb, Se, and Zn. Seven laboratories also calculated a number of hair "element ratios" from measured element concentrations (eg, sodium/potassium).

Some differences were found in laboratory sample preparation methods that could contribute to observed differences in results. Sample washing methods varied greatly among laboratories, including laboratory A that does not wash hair samples at all. Laboratories used either (plasma-coupled) atomic emission or mass spectroscopy for multiple element determination (Table 1). The potential advantage of mass spectroscopy is its lower limits of detection.⁴ This may account for part of the variation in detection limits, for example, lithium in Table 2 and Table 3. Laboratories varied in their calibration standards. Some laboratories used only repeat analysis of in-house pooled hair samples. Others purchased nonhair mineral standards (for example, aqueous standards). Several laboratories claimed the use of a Chinese commercial hair standard certified for up to 17 elements.⁵ At the time of this report, there were no hair standards certified for all elements analyzed by the laboratories. Additionally, certified values of several elements in this commercial hair standard exceed, sometimes by several-fold, the laboratories' own reference ranges.⁵

All laboratories identified a normal reference range for each analyzed element. Table 2 summarizes the spread of reference ranges across laboratories for the most commonly analyzed elements. In telephone interviews, 5 of the 6 selected laboratories indicated their reference ranges represented concentrations found in approximately two thirds of their population, consistent with typical clinical reporting methods. Laboratory A, the sixth selected laboratory, indicated that its reference ranges also were based on an older (unspecified) study of "healthy athletes." As noted in Table 2, interlaboratory differences in several cases are wide enough that the reference ranges do not overlap.

Table 3 shows results of the split hair sample submitted to the 6 selected laboratories, for elements analyzed by 3 or more laboratories. Many reported values showed great differences. Twelve elements with nonzero reported values have an order of magnitude or greater difference between the highest and lowest values (ie, As, Ba, B, Co, Ge, Hg, Mn, Mo, Pb, Na, Ni, and Sn). Outliers were identified in 14 elements' high value(s) as statistically significant (P<.05) extreme values compared with the same element from the other laboratories. Laboratory B reported extreme high values for 13 elements, and laboratories D and E each reported 2 elements with extreme high values. No low values tested as significantly extreme. Focusing on the 19 elements reported by all selected laboratories, the nonparametric Friedman analysis of variance indicated a significant difference across laboratories (P<.001), primarily because of laboratory B's extreme high values.

Each of the 6 selected laboratories rated their findings as above, within, or below normal range. A comparison of their findings for the 19 shared elements is shown in Table 4. Five of the 6 laboratories reported at least 1 element above the normal range, but none flagged the same element. Laboratory A reported aluminum above the normal range; laboratory B, phosphorus; laboratory C, arsenic and lead; laboratory D, molybdenum; and laboratory E, mercury.

Fourteen of the 19 elements were reported as below the normal range by at least 1 laboratory. Here the reporting was more consistent. All 6 laboratories reported manganese below the normal range. There was disagreement whether 2 elements were above or below the normal range: molybdenum was reported as above the normal range by laboratory D, but below by laboratories A and E; phosphorus was above the normal range by laboratory B but below by laboratory A. Table 4 also reveals a propensity for laboratory A to flag elements as deficient. Fourteen of the 19 elements were classified by laboratory A as below the normal range.

Table 5 summarizes laboratory interpretations of donor hair analyses. There was little agreement among laboratories as to which element concentrations and ratios were markers of disease. For example, 1 laboratory placed heavy diagnostic emphasis on hair sodium measurements, while 2 other laboratories dismissed hair sodium measurements as unreliable. Laboratory A assigned "high" or "low" classifications for concentrations interpreted as "non-ideal," even for those within the reference range (eg, zinc).

Four laboratories recommended vitamin/mineral supplementation of which 3 (A, E, and F) recommended a proprietary product line costing up to $100 per month for an indeterminate duration. Conflicting interpretation and dietary recommendations were observed between laboratories. For example, laboratory E identified the patient as a "fast metabolizer" and recommended a dietary increase in purine-containing protein and dairy products and abstention from vitamin A supplementation. Laboratory A identified the patient as a "slow metabolizer" and included vitamin A in its supplement recommendations, while advising avoidance of high-purine proteins and dairy products.

It is easy to understand the practical and entrepreneurial appeal of hair analysis to assess toxic exposure. Hair is a tissue easily obtained without pain or risk. It grows 1 to 2 cm per month and integrates biochemical events at the hair follicle over its entire length. Hair analysis has been used to confirm human poisonings by some elements, notably mercury and arsenic.^{6 - 7} But for most other substances there is little or no experimental evidence supporting hair as a true biological marker.^{8 - 11} Very little information exists regarding the physiological relationship between the biological uptake of minerals or chemicals and the concentrations delivered to the hair follicle, which are then incorporated into hair.^{11 - 14} Although hair serves as a minor excretory organ, the relationship between element concentrations in the blood compartment and those in hair is not a simple one.^{12 ,15 - 16} Wide variations in normal concentrations of most minerals in hair, due in part to differences in age, race, sex, geographic location, hair color, and hair treatment,^{17 - 23} complicate the clinical and epidemiological use of the analysis.^{10 - 11 ,24 - 31} Methylmercury is the only chemical with an established threshold for target organ effects that is based on hair element concentrations.^{6 ,32} There is no reliable laboratory method for separating external contaminants from biologically incorporated elements, and washing procedures can alter the concentrations of many elements in hair.^{12 ,17 ,33 - 38} Even for methylmercury, factors such as cosmetic hair treatments may diminish the use of hair as an accurate exposure biomarker.¹⁹

This study assessed reliability of hair trace element measurements across laboratories. The primary limitation was the amount of uniform scalp hair sample available for analysis. Obtaining scalp hair from 1 donor at 1 point in time was a crucial aspect of the study design. This necessarily limited the selection of laboratories receiving a split of the sample to 6 of the 9 identified US laboratories. It also was not possible to assess intralaboratory variability through the use of duplicate samples sent to each laboratory.

Many laboratories offering hair analysis advertise CLIA certification as evidence of their commitment to quality control. HCFA staff confirmed that hair mineral analysis falls under CLIA regulation as a "high complexity" test. Certification is done by specialty and subspecialty (such as chemistry or toxicology), not by specific analytes. Laboratories performing high complexity tests are required to participate in proficiency testing (PT) programs.³⁹ Laboratories conducting tests for which PT is not available are required to have a system for verifying the accuracy of their test results at least twice a year. However, an approved PT program for hair mineral analysis does not exist. The verification method and criteria for accuracy are left up to each laboratory. It would seem that the absence of available PT should engender more regulatory oversight, not less.

We found significant interlaboratory differences in reference ranges, hair test values, and interpretations consistent with previous studies.^{1 ,4} The most inconsistent results were the laboratory interpretation of the hair element values. Sources of discrepant interpretation included different reference ranges and the generation of element ratios, which are purported to provide additional information regarding health effects, such as altered energy levels or glucose/carbohydrate sensitivity. Laboratories did not provide evidence to justify the use of these ratios other than their own internal data. A MEDLINE search of the scientific peer-reviewed literature provided no information regarding this practice.

There was no agreement across laboratories in characterizing toxic concentrations of metals. Arsenic, mercury, or aluminum concentrations were labeled as toxic by single laboratories, with warnings of environmental sources and suggestions for retesting. Laboratories indicated that our test donor was at increased risk for various adverse health effects, including adrenal insufficiency, anemia, cardiovascular disease, dysinsulinism, passive-aggressive behavior patterns, and others. There was no clinical evidence that the donor had such health problems.

Of the 13 laboratories in the 1985 study by Barrett¹ (written communication identifying laboratories from S. Barrett, MD, June 12, 2000), 3 were still performing hair mineral analyses. Since 1985, 6 additional US laboratories have begun offering these analyses. Despite the passage of the CLIA in 1988, hair analysis laboratories are thriving, and there are no federally recognized PT reference laboratories for hair element/mineral analyses. Increasingly, members of the public have access to Internet sites that extol the benefits of this procedure and provide access to testing with or without involving a personal physician.

Hair analysis of individuals for trace elements and nutritional balance is generally unreliable. This is because (1) many factors, such as hair treatment and external contamination, can affect hair mineral concentrations; (2) even when these variables are standardized, as in our study, analytical variability among commercial laboratories conducting nutritional hair mineral analyses makes interpretation of such data problematic; and (3) dose-response data linking hair mineral concentrations to target organ effects are largely unavailable. Health care choices based on these analyses may be ineffective or even detrimental to the patient's overall health. We recommend the following: (1) Primary care clinicians refrain from using hair analysis to assess environmental exposures or nutritional balance. (2) Public health and consumer protection agencies provide warnings to the public about the general unreliability of these tests. (3) HCFA refrains from certifying laboratories for hair analysis until standards for PT are developed. (4) HCFA prohibits the display of CLIA certification information on hair analysis advertising and submittal forms for laboratories conducting hair analyses when such certification applies to other sample media. (5) The US Food and Drug Administration or Bureau of Consumer Protection, Federal Trade Commmission evaluates laboratory claims based on hair analysis, particularly those suggesting medical abnormalities requiring additional laboratory diagnostics, nutritional supplements, or major changes in dietary habits.